Isabella and Catalina

[Frank21]

Circa 1594 princesses Isabella and Catalina of Spain, and I feel somewhat emboldened to post some SD images after so many of us have come out of the SD closet in the art comp 😴. There's some clipping in the highlights but the people and clothing look real compared to DS. Perhaps SD has spoiled me but when I look at the Daz store I shudder at how amateurish the (characters/clothing) everything looks now.

[full image]

[Serge2804]

for sure this image looks real, closed a photo.
When you say "SD" do you talk about Poser and "DS" daz ?

[Frank21]

Thanks Serge, DS = Daz Studio, and SD = Stable Diffusion. It's a pity the two acronyms are so similar when they both relate to art.

[Architector]

Very nice results and moody portrait. I love how the ai create younger shapes from adults.

[Legolas18]

Very nice! The hands came out really well, and Jenna Ortega seems to work well here 🙂

I usually don't go for photographic style, but at least SD offers literally hundreds of styles to choose from, plus you can mix styles for almost endless combinations.

As for the Daz characters you mentioned...I agree, after my Daz hiatus, even the best of them look just "meh". Plus they are getting more and more expensive, so I am just keeping poses/expressions to use with ControlNet.

[gaver]

It looks very slick.
In the SD/DS discussion....
I guess it is just what you like to accomplish.
I like the , sometimes, almost " drawn" effect I get with DS
I know it is limited compared to SD, but that is ok.
As I said before, I like the creative process, but once rendered, it is not more then the next pic in my desktop slideshow.
And I like to see what I do as I get along.
What I understand of AI art, is you just add a lot of keywords and press enter.
Though the results are great, it is not my way of doing things

[Legolas18]

@gaver SD can be just about "putting in a lot words" or much more complex, like using Daz poses and expressions to control characters, light maps to control light, photo collages to control composition, or even creating your own node-based image processing setup.

[gaver]

@Legolas I am sure that it is more than what I wrote. Can;t get my head around this whole Ai thiong. Maybey it is just I am getting to old to learn new tricks..

[Frank21]

It took me the best part of 6 months before I knew what I was doing gaver and I'm probably your age or older. Just getting two different people into a scene is hugely problematic.
The pic above actually has the same characters but they look different because they are different ages, so I called them sisters (cheat). To get to that stage took several hours of refinement and a bit of postwork. The whole process needs patience and can be really infuriating.

[diviniren]

@frank22 been a while! These two look great. Your SD skills have come a long way since when you first started! Only thing I'm noticing is this theme of high gloss and noise reduction on most SD outputs involving people/skin. Is there a way to reduce that and add back detail?

[Frank21]

Thanks diviniren.
A lot of people run their images through "Extras" for image enlargement. The side effect of this is that the image becomes a bit blurry, contrasty [sic] and has an air-brushed look. My 2080 can render txt2img images up to about 2000x1400 which is big enough for me so I avoid using extras.

[Author]

Posts